This project is a complete Vietnamese OCR pipeline that combines object detection and sequence modeling to detect and recognize text in images. The system uses YOLOv5 for text region detection and VietOCR (Transformer-based) for recognizing the text content.
You can watch the video demo here
The pipeline consists of two main stages:
- Text Detection (YOLOv5):
- We use YOLOv5 as the text detector.
- It is a fast and accurate real-time object detection algorithm.
- The model is trained (or fine-tuned) to detect text areas in input images and return bounding boxes around them.
- These cropped regions are then passed to the next stage for recognition.
Figure: Detection results
- Text Recognition (VietOCR):
- We use VietOCR, a deep learning-based text recognizer optimized for Vietnamese.
- VietOCR architecture consists of two main components:
- VGG-style CNN for feature extraction: Converts input images into visual feature sequences.
- Transformer encoder-decoder: Translates the visual features into Vietnamese character sequences.
- Unlike traditional OCRs using CTC, VietOCR uses Seq2Seq with attention, making it highly accurate for Vietnamese with diacritics and complex scripts.
Figure: VietOCR results
Feel free to open issues or pull requests if you'd like to improve this project or add new features.
This project is open-source and available under the MIT License.
Thank you for checking out this project! If you find it useful, feel free to ⭐️ star the repository and share it with others.